•Deep-learning networks for prediction of PBS and VMAT doses were established.•Relative errors between planned and predicted doses were minimal in the target.•Strong correlation (r > 0.9) was found ...between planned and predicted DVH metrics.•The networks classified the PBS dosimetric benefit over VMAT with 98% precision.•Proposed networks can accurately predict the treatment modality selection outcome.
Dose prediction using deep learning networks prior to radiotherapy might lead tomore efficient modality selections. The study goal was to predict proton and photon dose distributions based on the patient-specific anatomy and to assess their clinical usage for paediatric abdominal tumours.
Data from 80 patients with neuroblastoma or Wilms’ tumour was included. Pencil beam scanning (PBS) (5 mm/ 3%) and volumetric-modulated arc therapy (VMAT) plans (5 mm) were robustly optimized on the internal target volume (ITV). Separate 3-dimensional patch-based U-net networks were trained to predict PBS and VMAT dose distributions. Doses, planning-computed tomography images and relevant optimization masks (ITV, vertebra and organs-at-risk) of 60 patients were used for training with a 5-fold cross validation.
The networks’ performance was evaluated by computing the relative error between planned and predicted dose-volume histogram (DVH) parameters for 20 inference patients. In addition, the organs-at-risk mean dose difference between modalities was calculated using planned and predicted dose distributions (ΔDmean = DVMAT-DPBS). Two radiation oncologists performed a blind PBS/VMAT modality selection based on either planned or predicted ΔDmean.
Average DVH differences between planned and predicted dose distributions were ≤ |6%| for both modalities. The networks classified the organs-at-risk Dmean difference as a gain (ΔDmean > 0) with 98% precision. An identical modality selection based on planned compared to predicted ΔDmean was made for 18/20 patients.
Deep learning networks for accurate prediction of proton and photon dose distributions for abdominal paediatric tumours were established. These networks allowing fast dose visualisation might aid in identifying the optimal radiotherapy technique when experience and/or resources are unavailable.
The integration of real-time magnetic resonance imaging (MRI) guidance and proton therapy would potentially improve the proton dose steering capability by reducing daily uncertainties due to ...anatomical variations. The use of a fixed beamline coupled with an axial patient couch rotation would greatly simplify the proton delivery with MRI guidance. Nonetheless, it is mandatory to assure that the plan quality is not deteriorated by the anatomical deformations due to patient rotation. In this work, an in-house tool allowing for intra-fractional per-beam adaptation of intensity-modulated proton plans (BeamAdapt) was implemented through features available in RayStation. A set of three MRIs was acquired for two healthy volunteers (
,
): (1) no rotation/static, (2) rotation to the right and (3) left.
was rotated by 15°, to simulate a clinical pediatric abdominal case and
by 45°, to simulate an extreme patient rotation case. For each volunteer, a total of four intensity-modulated pencil beam scanning plans were optimized on the static MRI using virtual abdominal targets and two-three posterior-oblique beams. Beam angles were defined according to the angulations on the rotated MRIs. With BeamAdapt, each original plan was initially converted into separate plans with one beam per plan. In an iterative order, individual beam doses were non-rigidly deformed to the rotated anatomies and re-optimized accounting for the consequent deformations and the beam doses delivered so far. For evaluation, the final accumulated dose distribution was propagated back to the static MRI. Planned and adapted dose distributions were compared by computing relative differences between dose-volume histogram metrics. Absolute target dose differences were on average below 1% and organs-at-risk mean dose differences were below 3%. With BeamAdapt, not only intra-fractional per-beam proton plan adaptation coupled with axial patient rotation is possible but also the need for a rotating gantry during MRI guidance might be mitigated.
Abstract Background and purpose A methodology is presented to quantify the uncertainty associated with linear accelerator-based frameless intracranial stereotactic radiotherapy (SRT) combining ...end-to-end phantom tests and clinical data. Methods and materials The following steps of the SRT chain were analysed: planning computed tomography (CT) and magnetic resonance (MR) scans registration, target volume delineation, CT and cone beam CT (CBCT) registration and intrafraction-patient displacement. The overall accuracy was established with an end-to-end test. The measured uncertainties were combined, deriving the total systematic ( ΣT ) and random ( σT ) error components, to estimate the GTV-PTV margin. Results The uncertainty in the MR-CT registration was on average 0.40 mm (averaged over AP, CC and LR directions). Rotational variations were smaller than 0.5° in all directions. Interobser variation in GTV delineation was on average 0.29 mm. The uncertainty in the CBCT-CT registration was on average 0.15 mm. Again, rotational variations were smaller than 0.5° in all directions. The systematic and random intrafraction displacement errors were on average 0.55 mm and 0.45 mm, respectively. The systematic and random positional errors from the end-to-end test were on average 0.49 mm and 0.53 mm, respectively. Combining these uncertainties resulted in an average ΣT = 0.9 mm and σT = 0.7 mm and an average GTV-PTV margin of 2.8 mm. Conclusion This comprehensive methodology including end-to-end tests enabled a GTV-PTV margin calculation considering all sources of uncertainties. This generic method can also be used for other treatment sites.
Positron emission tomography (PET) is a promising tool for monitoring the three-dimensional dose distribution in charged particle radiotherapy. PET imaging during or shortly after proton treatment is ...based on the detection of annihilation photons following the ß(+)-decay of radionuclides resulting from nuclear reactions in the irradiated tissue. Therapy monitoring is achieved by comparing the measured spatial distribution of irradiation-induced ß(+)-activity with the predicted distribution based on the treatment plan. The accuracy of the calculated distribution depends on the correctness of the computational models, implemented in the employed Monte Carlo (MC) codes that describe the interactions of the charged particle beam with matter and the production of radionuclides and secondary particles. However, no well-established theoretical models exist for predicting the nuclear interactions and so phenomenological models are typically used based on parameters derived from experimental data. Unfortunately, the experimental data presently available are insufficient to validate such phenomenological hadronic interaction models. Hence, a comparison among the models used by the different MC packages is desirable. In this work, starting from a common geometry, we compare the performances of MCNPX, GATE and PHITS MC codes in predicting the amount and spatial distribution of proton-induced activity, at therapeutic energies, to the already experimentally validated PET modelling based on the FLUKA MC code. In particular, we show how the amount of ß(+)-emitters produced in tissue-like media depends on the physics model and cross-sectional data used to describe the proton nuclear interactions, thus calling for future experimental campaigns aiming at supporting improvements of MC modelling for clinical application of PET monitoring.
To perform patient plan quality assurance (QA) on a newly installed MR-linac (MRL) it is necessary to have an MR-compatible QA device. An MR compatible device (MR-Delta4) has been developed together ...with Scandidos AB (Uppsala, Sweden). The basic characteristics of the detector response, such as short-term reproducibility, dose linearity, field size dependency, dose rate dependency, dose-per-pulse dependency and angular dependency, were investigated for the clinical Delta4-PT as well as for the MR compatible version. All tests were performed with both devices on a conventional linac and the MR compatible device was tested on the MRL as well. No statistically significant differences were found in the short-term reproducibility (<0.1%), dose linearity (⩽0.5%), field size dependency (<2.0% for field sizes larger than 5 × 5 cm
), dose rate dependency (<1.0%) or angular dependency for any phantom/linac combination. The dose-per-pulse dependency (<0.8%) was found to be significantly different between the two devices. This difference can be explained by the fact that the diodes in the clinical Delta4-PT were irradiated with a much larger dose than the MR-Delta4-PT ones. The absolute difference between the devices (<0.5%) was found to be small, so no clinical impact is expected. For both devices, the results were consistent with the characteristics of the Delta4-PT device reported in the literature (Bedford et al 2009 Phys. Med. Biol. 54 N167-76; Sadagopan et al 2009 J. Appl. Clin. Med. Phys. 10 2928). We found that the characteristics of the MR compatible Delta4 phantom were found to be comparable to the clinically used one. Also, the found characteristics do not differ from the previously reported characteristics of the commercially available non-MR compatible Delta4-PT phantom. Therefore, the MR compatible Delta4 prototype was found to be safe and effective for use in the 1.5 tesla magnetic field of the Elekta MR-linac.
•The entire brain is susceptible to dose-dependent volume loss after RT.•Future studies should examine the impact of cerebral volume loss on cognition.•Current sparing strategies in RT for brain ...tumours may need to be reconsidered.
Numerous brain MR imaging studies have been performed to understand radiation-induced cognitive decline. However, many of them focus on a single region of interest, e.g. cerebral cortex or hippocampus. In this study, we use deformation-based morphometry (DBM) and voxel-based morphometry (VBM) to measure the morphological changes in patients receiving fractionated photon RT, and relate these to the dose. Additionally, we study tissue specific volume changes in white matter (WM), grey matter (GM), cerebrospinal fluid and total intracranial volume (TIV).
From our database, we selected 28 patients with MRI of high quality available at baseline and 1 year after RT. Scans were rigidly registered to each other, and to the planning CT and dose file. We used DBM to study non-tissue-specific volumetric changes, and VBM to study volume loss in grey matter. Observed changes were then related to the applied radiation dose (in EQD2). Additionally, brain tissue was segmented into WM, GM and cerebrospinal fluid, and changes in these volumes and TIV were tested.
Performing DBM resulted in clusters of dose-dependent volume loss 1 year after RT seen throughout the brain. Both WM and GM were affected; within the latter both cerebral cortex and subcortical nuclei show volume loss. Volume loss rates ranging from 5.3 to 15.3%/30 Gy were seen in the cerebral cortical regions in which more than 40% of voxels were affected. In VBM, similar loss rates were seen in the cortex and nuclei. The total volume of WM and GM significantly decreased with rates of 5.8% and 2.1%, while TIV remained unchanged as expected.
Radiotherapy is associated with dose-dependent intracranial morphological changes throughout the entire brain. Therefore, we will consider to revise sparing of organs at risk based on future cognitive and neurofunctional data.
Purpose/Objective: Flank irradiation for Wilms' tumor (WT) is currently performed at our institute using a cone-beam computed tomography-guided volumetric modulated arc (VMAT
CBCT
) workflow. By ...adding real-time magnetic resonance imaging (MRI) guidance to the treatment, safety margins could be reduced. The study purpose was to quantify the potential reduction of the planning target volume (PTV) margin and its dosimetric impact when using an MRI-guided intensity modulated radiation therapy (IMRT
MRI
) workflow compared to the VMAT
CBCT
workflow.
Material/Methods: 4D-CT, MRI and CBCT scans acquired during preparation and treatment of 15 patients, were used to estimate both geometric, motion and patient set-up systematic (∑) and random (σ) errors for VMAT
CBCT
and IMRT
MRI
workflows. The mean PTV (PTV
mean
) expansion was calculated using the van Herk formula. Treatment plans were generated using five margin scenarios (PTV
mean
± 0, 1 and 2 mm). Furthermore, the IMRT
MRI
plans were optimized with a 1.5T transverse magnetic field turned-on to realistically model an MRI-guided treatment. Plans were evaluated using dose-volume statistics (p<.01, Wilcoxon).
Results: Analysis of ∑ and σ errors resulted in a PTV
mean
of 5 mm for the VMAT
CBCT
and 3 mm for the IMRT
MRI
workflows in each orthogonal direction. Target coverage was unaffected by the margin decrease with a mean V
95%
=100% for all margin scenarios. For the PTV
mean
, an average reduction of the mean dose to the organs at risk (OARs) was achieved with IMRT
MRI
compared to VMAT
CBCT
: 3.4 ± 2.4% (p<.01) for the kidney, 3.4 ± 2.1% (p<.01) for the liver, 2.8 ± 3.0% (p<.01) for the spleen and 4.9 ± 3.8% (p<.01) for the pancreas, respectively.
Conclusions: Imaging data in children with WT demonstrated that the PTV margin could be reduced isotropically down to 2 mm when using the IMRT
MRI
compared to the VMAT
CBCT
workflow. The former results in a dose reduction to the OARs while maintaining target coverage.
The use of Stereotactic Body Radiotherapy (SBRT) for bone metastases is increasing rapidly. Therefore, knowledge of the inter-observer differences in tumor volume delineation is essential to ...guarantee precise dose delivery. The aim of this study is to compare inter-observer agreement in bone metastases delineated on different imaging modalities.
Twenty consecutive patients with bone metastases treated with SBRT were selected. All patients received CT and MR imaging in treatment position prior to SBRT. Five observers from three institutions independently delineated gross tumor volume (GTV) on CT alone, CT with co-registered MRI and MRI alone. Four contours per imaging modality per patient were available, as one set of contours was shared by 2 observers. Inter-observer agreement, expressed in generalized conformity index CIgen, volumes of contours and contours center of mass (COM) were calculated per patient and imaging modality.
Mean GTV delineated on MR (45.9±52.0cm3) was significantly larger compared to CT–MR (40.2±49.4cm3) and CT (34.8±41.8cm3). A considerable variation in CIgen was found on CT (mean 0.46, range 0.15–0.75) and CT–MRI (mean 0.54, range 0.17–0.71). The highest agreement was found on MRI (mean 0.56, range 0.20–0.77). The largest variations of COM were found in anterior–posterior direction for all imaging modalities.
Large inter-observer variation in GTV delineation exists for CT, CT–MRI and MRI. MRI-based GTV delineation resulted in larger volumes and highest consistency between observers.